Explore · 每周精选
发现最受关注的研究论文,追踪研究趋势,订阅感兴趣的期刊与关键词。
50 篇论文 · 按点赞排序
Josef Dai, Xuehai Pan, Ruiyang Sun +5 authors
Safe RLHF, a novel algorithm for human value alignment in LLMs, improves performance and safety by decoupling human preferences and using a Lagrangian method to balance reward and cost constraints.
Elias Frantar, Dan Alistarh
QMoE enables efficient execution of trillion-parameter MoE models on affordable hardware through compression and bespoke GPU decoding.
Fuxiao Liu, Tianrui Guan, Zongxia Li +4 authors
HallusionBench is a benchmark that highlights language hallucination and visual illusion issues in vision-language models (VLMs), showcasing the limitations of current state-of-the-art models like GPT-4V and LLaVA-1.5.
Yecheng Jason Ma, William Liang, Guanzhi Wang +6 authors
Eureka, an LLM-powered reward design algorithm, generates human-level reward functions enhancing reinforcement learning for complex manipulation tasks and RLHF, including pen spinning with a simulated Shadow Hand.
Joey Hejna, Rafael Rafailov, Harshit Sikchi +4 authors
A new regret-based algorithm, Contrastive Preference Learning (CPL), learns optimal policies directly from human preferences without learning a reward function, addressing optimization challenges in Reinforcement Learning from Human Feedback (RLHF).
Tianyi Chen, Tianyu Ding, Badal Yadav +2 authors
LoRAShear is an efficient method for pruning and knowledge recovery in Large Language Models, reducing their size by 20% with minimal performance loss.
Haoxiang Wang, Pavan Kumar Anasosalu Vasu, Fartash Faghri +6 authors
A unified model combining CLIP and SAM using multi-task learning, continual learning, and distillation techniques achieves state-of-the-art zero-shot semantic segmentation performance with reduced computational and data requirements.
Dingyao Yu, Kaitao Song, Peiling Lu +5 authors
MusicAgent is a system that integrates various music-related tools and uses large language models to automate music processing tasks based on user requests.
Ziqi Wang, Le Hou, Tianjian Lu +4 authors
A new framework, ImPlicit Self-ImprovemenT (PIT), uses human preference data to implicitly enhance the quality of Large Language Model responses, outperforming prompting-based methods.
Yang Wu, Shilong Wang, Hao Yang +4 authors
GPT-4V demonstrates strong visual understanding but has limitations in language comprehension, handling sensitive data, modalities like depth and audio, and fine visual nuances.
Xiaoxiao Long, Yuan-Chen Guo, Cheng Lin +8 authors
Wonder3D leverages cross-domain diffusion models and geometry-aware normal fusion for high-quality, efficient, and consistent 3D mesh generation from single-view images.
Jun Chen, Deyao Zhu, Xiaoqian Shen +7 authors
MiniGPT-v2 is a unified vision-language model that uses task-specific identifiers to perform diverse tasks with high performance.
Mihir Prabhudesai, Anirudh Goyal, Deepak Pathak +1 authors
AlignProp refines text-to-image diffusion models using backpropagation through the denoising process, leveraging low-rank adapters and gradient checkpointing to optimize for various objectives with higher efficiency.
Samuel L. Smith, Andrew Brock, Leonard Berrada +1 authors
ConvNets pre-trained on a large dataset match the performance of Vision Transformers on ImageNet with comparable computational resources.
Kevin Lin, Faisal Ahmed, Linjie Li +9 authors
MM-VID integrates GPT-4V with specialized tools to transcribe and understand long-form, multimodal videos, supporting advanced tasks like audio description and character identification.
Tu Vu, Mohit Iyyer, Xuezhi Wang +8 authors
FreshQA evaluates LLMs' factuality and hallucination in responding to current knowledge questions, revealing limitations and demonstrating improvements through FreshPrompt, an augmented prompting method using search engine data.
Dongchao Yang, Jinchuan Tian, Xu Tan +9 authors
UniAudio, a unified language model system, generates various audio types effectively using tokenization, sequence concatenation, and next-token prediction, with a multi-scale Transformer model handling extended sequences.
Ruida Wang, Wangchunshu Zhou, Mrinmaya Sachan
Synthesis Step by Step (S3) iteratively refines synthesized pseudo training examples using a large language model to reduce distributional discrepancies and improve small model performance on NLP tasks.
Juan Rocamonde, Victoriano Montesinos, Elvis Nava +2 authors
Pretrained vision-language models can serve as zero-shot reward models in reinforcement learning, enabling complex task learning from minimal natural language prompts.
Kangfu Mei, Mauricio Delbracio, Hossein Talebi +3 authors
A novel single-stage distillation method for generative diffusion models reduces sampling time while maintaining performance across tasks like super-resolution and image editing.
北京市昌平区探索星信息技术及软件开发工作室
京ICP备2026059466号